Abstract: The method of state iteration and parallel computation for a class of the generalized linear quadric optimal, control models is presented. The method devised is effective in overcoming the dimensionality difficulties in multi-state dynamic decision problems and can greatly accelerate the speed of computation with parallel computers.For the stated model, this paper gives the relevant theoretical results. The paper corrects a key error in the results given by some authors (Ref. 1) and extends their results.